無料GitHub で公開

squad-plan

Generate a parallel workstream plan for multi-agent Claude Code work. Decomposes a feature into independent workstreams, each ownable by a separate Claude Code instance. Saves to claude/tasks/ACTION_PLAN.md.

日本語の概要は準備中です。原文の説明を表示しています。

brain-bootstrap/claude-code-brain-bootstrap

実行スコア未検証

インストール方法を見る

含まれるファイル（1）

SKILL.md2.6 KB

SKILL.md（原文）

インストールする前に、エージェントに与えられる指示の中身を確認できます。

Squad Plan Skill

Generate a detailed parallel execution plan for multi-agent work on a complex feature.

When to Use

Task touches ≥3 independent components/files
Components don't share mutable state during the task
Estimated single-agent time > 15 minutes
Each subtask can own a distinct file set (no overlap)

Protocol

Phase 1: Analyze the Task

Read CLAUDE.md and claude/architecture.md for project context
Break down the task into independent subtasks
Identify natural parallelism boundaries (frontend/backend, services, features, layers)
Map dependencies — what must complete before something else starts
Identify shared resources requiring coordination (DB schemas, API contracts, shared types)

Phase 2: Design Workstreams (2-5 workstreams)

For each workstream:

Workstream ID & Title — e.g., "WS1: API Layer"
Objective — what this workstream accomplishes
Files/Components owned — specific parts of the codebase
Tasks — numbered checklist items with expected outcomes and verification steps
Dependencies — which tasks from other workstreams must complete first
Deliverables — what this workstream produces for others to consume

Phase 3: Integration Strategy

Define integration points between workstreams
Create integration test checklist
Specify merge order (which branch merges first)
Define smoke test for the combined result

Phase 4: Write ACTION_PLAN.md

Save to claude/tasks/ACTION_PLAN.md:

# ACTION_PLAN — [Feature Name]

## Overall Objective
[Clear statement of the goal]

## Workstream Definitions

### WS1: [Title]
**Objective:** [Goal]
**Files:** [List]

#### Tasks
- [ ] Task 1: [Description]
- [ ] Task 2: ...

**Dependencies:** None (can start immediately)
**Delivers to:** WS2 needs [output] from Task 3

### WS2: [Title]
...

## Integration Points
- [ ] [Integration task 1]

## Testing Strategy
- [ ] Unit tests per workstream
- [ ] Integration tests after merge
- [ ] End-to-end smoke test

Usage with Multiple Claude Code Instances

After generating, assign instances:

"You are the Backend Developer. Execute Workstream WS1 from claude/tasks/ACTION_PLAN.md."
"You are the Frontend Developer. Execute WS2, noting you depend on WS1 Task 3."

Each instance reads CLAUDE.md for project standards and follows the Explore-Plan-Act workflow.

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

ログインなしで書き込めます。使ってみた感想、質問、動かなかった環境などを気軽にどうぞ。

まだコメントはありません。最初のコメントを書いてみませんか。

同じリポジトリのスキル

ask

無料

Route a codebase question to the right tool — structural graph, semantic search, or risk analysis. Use when you need to find or understand code without knowing exact file names.

日本語の概要は準備中です。原文の説明を表示しています。